Depth Quality-Inspired Feature Manipulation for Efficient RGB-D Salient Object Detection

Abstract

RGB-D salient object detection (SOD) recently has attracted increasing research interest by benefiting conventional RGB SOD with extra depth information. However, existing RGB-D SOD models often fail to perform well in terms of both efficiency and accuracy, which hinders their potential applications on mobile devices and real-world problems. An underlying challenge is that the model accuracy usually degrades when the model is simplified to have few parameters. To tackle this dilemma and also inspired by the fact that depth quality is a key factor influencing the accuracy, we propose a novel depth quality-inspired feature manipulation (DQFM) process, which is efficient itself and can serve as a gating mechanism for filtering depth features to greatly boost the accuracy. DQFM resorts to the alignment of low-level RGB and depth features, as well as holistic attention of the depth stream to explicitly control and enhance cross-modal fusion. We embed DQFM to obtain an efficient light-weight model called DFM-Net, where we also design a tailored depth backbone and a two-stage decoder for further efficiency consideration. Extensive experimental results demonstrate that our DFM-Net achieves state-of-the-art accuracy when comparing to existing non-efficient models, and meanwhile runs at 140ms on CPU (2.2× faster than the prior fastest efficient model) with only ∼8.5Mb model size (14.9% of the prior lightest). Our code will be available at https://github.com/zwbx/DFM-Net. CCS CONCEPTS• Computing methodologies → Interest point and salient region detections.

EINet: camouflaged object detection with pyramid vision transformer

“…To underline the importance of texture-aware features for COD, Ren et al 11 presented a refinement module based on deep texture perception. Zhang et al 26 suggested a confidence-aware COD framework to prevent deep networks from being overconfident. In the latest research, Pang et al 27 proposed a mixed-scale triplet state network, namely ZoomNet.…”

Multi-Modality Diversity Fusion Network with Swintransformer for RGB-D Salient Object Detection

“…To testify the validity of our model, we conduct comprehensive experiments on five datasets with 17 SOTA methods, including A2dele [6], BiANet [10], CDINet [7], CMWNet [14], CoNet [8], D3Net [9], DANet [5], DCFNet [13], DFMNet [15], DSNet [4], DSA2F [16], HDFNet [11], ICNet [17], JL-DCF [18], SMEG [19], SSF [20], and UC-Net [21]. Quantitative comparison: Table 1 shows the quantitative results of the M 2 DFNet and other methods on five datasets.…”

Bi-Directional Progressive Guidance Network for RGB-D Salient Object Detection

“…Comparison with the State-of-the-Arts 1) Comparison Methods: We compare our model with 13 state-of-the-art RGB-D based SOD methods, including MMCI [14],TANet [15], DMRA [18], ICNet [48], A2dele [66], S2MA [67], DRLF [38], CCAFNet [68], JL-DCF [41], CPFP [19], D3Net [13], DQSD [21] and DFMNet [69]. Specially, the qualities of depth images are also considered in the last three methods, i.e., CPFP [19], D3Net [13], DQSD [21] and DFMNet [69]. For all the methods mentioned here, we use either the implementations with their default parameter settings or the saliency maps provided by their authors for fair comparisons.…”